Research On Single-stage Detection Algorithm For Fire Smok

As global temperatures rise,the frequency of fires is increasing year by year,posing a serious threat to human life and property and the natural environment on which they depend.The key to reducing fire hazards is to detect and identify smoke in the early stages of a fire,identify fires,and handle them.However,traditional smoke detectors can be affected by the surrounding environment,resulting in slow detection speed,false detection and missed detection,as well as poor detection performance when multiple targets exist in a single image,which makes it difficult to achieve the desired detection performance.In response to the above issues,this article has conducted research on smoke detection technology in the early stage of a fire,mainly optimizing the processing based on the YOLOv5 s algorithm,and proposed a YOLOv5s-RAS smoke detection model with faster speed,higher accuracy,and better performance.The main work contents are as follows:(1)Production data set: The data set mainly consists of two categories: flame and smoke.Based on the Baidu Feijiang fire smoke dataset,a dataset is established by using web crawler technology to obtain more images,and ensuring that each image has one or more interference items.Through data cleaning,20114 valid images are ultimately obtained,and they are divided into training sets,test sets,and verification sets.(2)To solve the problem that the complexity of network computing increases with the depth of the network,a YOLOv5s-R network is proposed.At the same time,a YOLOv5s-RA network is proposed to solve the problems of low smoke concentration in images that cannot be correctly identified and low accuracy of remote target detection.Compared to the original network,the average accuracy of the YOLOv5s-RA network,m AP,has been improved by 1.1%,the detection speed has been increased to 74 FPS,the parameter amount has been reduced by 52.1%,and the FLOPs have been reduced to 2.1G.The improved YOLOv5s-RA network significantly reduces the computational complexity,shortens the training time,and greatly improves the detection speed while ensuring that the detection accuracy does not decrease.(3)Although the detection speed of YOLOv5s-RA network has been greatly improved,its accuracy is relatively low,making it difficult to meet the high accuracy requirements of fire smoke detection.In order to solve the above problems,a sparrow search algorithm with strong optimization ability and fast convergence speed is used to optimize the model’s hyperparameters,and the optimized set of hyperparameters are substituted into the YOLOv5s-RA network.The experimental results show that compared with other networks,the ultimately optimized YOLOv5s-RAS network has a higher recognition effect on fire smoke.